Removal of Muscle Artifacts from Single-Channel EEG Based on Ensemble Empirical Mode Decomposition and Multiset Canonical Correlation Analysis

Xun Chen,Chen He,Hu Peng

doi:10.1155/2014/261347

Abstract

Electroencephalogram (EEG) recordings are often contaminated with muscle artifacts. This disturbing muscular activity strongly affects the visual analysis of EEG and impairs the results of EEG signal processing such as brain connectivity analysis. If multichannel EEG recordings are available, then there exist a considerable range of methods which can remove or to some extent suppress the distorting effect of such artifacts. Yet to our knowledge, there is no existing means to remove muscle artifacts from single-channel EEG recordings. Moreover, considering the recently increasing need for biomedical signal processing in ambulatory situations, it is crucially important to develop single-channel techniques. In this work, we propose a simple, yet effective method to achieve the muscle artifact removal from single-channel EEG, by combining ensemble empirical mode decomposition (EEMD) with multiset canonical correlation analysis (MCCA). We demonstrate the performance of the proposed method through numerical simulations and application to real EEG recordings contaminated with muscle artifacts. The proposed method can successfully remove muscle artifacts without altering the recorded underlying EEG activity. It is a promising tool for real-world biomedical signal processing applications.

Highlights

The electroencephalogram (EEG) is frequently contaminated by various physiological activities of noninterest, such as electrocardiogram (ECG), electrooculogram (EOG), and electromyogram (EMG)
While ECG and EOG artifacts can be effectively removed by using adaptive filters and blind source separation (BSS) techniques [1], the perturbation induced by muscular activity is difficult to correct as recently reviewed in [2]
The parameter τ in (12) can be chosen empirically as it may highly depend on the data structure

Summary

Introduction

The electroencephalogram (EEG) is frequently contaminated by various physiological activities of noninterest, such as electrocardiogram (ECG), electrooculogram (EOG), and electromyogram (EMG). These artifacts reduce the quality of the signal and blur features of interest. The main reason lies in the fact that EMG artifacts have higher amplitude (compared with the EEG signal), wide spectral distribution, and variable topographical distribution [2]. These muscle artifacts obscure EEG signals and make the interpretation of the EEG complicated or even unfeasible [3].

Methods

Results

Conclusion